Gargantua

Cooper lands on Miller's Planet. He's there for three hours and seventeen minutes. When he gets back to the Endurance, twenty-three years have passed. Romilly has aged. His children are grown. The messages have piled up. He sits in the ship and watches his kids become strangers on a screen, a decade per video, sobbing through the time he can never get back.

It's the most devastating scene in the film because it makes time dilation feel like something. Not an equation. A loss. The physics didn't betray Cooper. He understood the math. He knew what the black hole would cost. But understanding the ratio intellectually and experiencing the gap emotionally are different things.

I want to talk about why your project manager and an AI company are both doing the same thing Cooper did: staring at the math, realizing capacity is finite, and making the same hard choices about what to protect.

Because on March 26, 2026 — five days ago — Anthropic quietly adjusted the session limits for Claude Pro, Max, and Free users during peak hours. The token cost per session went up. The five-hour window started burning faster. Seven percent of users began hitting limits they'd never hit before. By today, March 31, Anthropic admitted that Claude Code quotas are "draining way faster than expected" and said it's their "top priority."

Their official recommendation? "Shift token-intensive background jobs to off-peak hours."

Read that again. An AI company — the company building the most capable AI models on the planet — just told its users to rearrange their work schedule around capacity constraints. To time-shift their demand. To self-triage. To accept that the system is finite and act accordingly.

Your project manager has been saying the same thing for twenty years. It's called "what's the priority?"

Section I: The Physics of Finite Capacity

In general relativity, time dilation occurs when a massive object curves spacetime. The closer you are to the mass, the slower time passes for you relative to a distant observer. On Miller's Planet, Gargantua's gravitational pull is so extreme that one hour on the surface equals seven years on the Endurance. Cooper isn't moving slowly — he's moving through time at a different rate than everyone else.

Now replace "massive object" with "AI-augmented team," and "spacetime" with "the velocity of work around you."

In our previous posts — Limitless: The Human Token Economy and Be Kind, Rewind — we established that a human knowledge worker produces roughly 4,000-8,000 tokens of meaningful work per day. Emails, Slack messages, documents, code, meeting contributions. Everything that leaves your brain through your fingers or your mouth.

With AI augmentation, your team's collective output isn't 5x that. It's 10-50x. The AI didn't just speed up individual contributors — it compressed timelines so dramatically that the gap between "what you produced" and "what happened around you" has become a time dilation problem.

You are Cooper. Your AI-augmented team is Gargantua. The closer you work to the team's AI-accelerated output, the more your personal time frame diverges from theirs. You step away for a meeting. You come back to a universe that aged weeks. Not metaphorically — in token output, the team produced what used to take weeks.

Section II: The Question Gap

Here's where the time dilation becomes personal. Think about how knowledge work actually flows. You don't work in isolation. You work in a cycle of doing and asking. You do work, you hit an ambiguity, you ask a question, you wait for the answer, you resume.

Every question you ask creates a gap. And in that gap, the world keeps moving.

Pre-AI, the gap was manageable. You post a question in Teams at 9am. By the time someone answers at 11am, roughly 2,000-4,000 tokens of work happened around you. Two hours of normal human output. You read the Slack thread. You're caught up in five minutes. No problem.

Post-AI, you post the same question at 9am. By 11am, your team — augmented by AI coding assistants, AI analysis tools, AI document generators — has produced 40,000-85,000 tokens. That's not a Slack thread you can skim. That's three design documents, two prototypes, a cost analysis, and a Slack thread with 47 messages debating the tradeoffs of an approach that didn't exist when you asked the question.

The question gap is the Miller's Planet of knowledge work. Every time you pause — to ask a question, to attend a meeting, to eat lunch, to think — you land on the surface. And when you come back, time has dilated.

This is why "scope creep" feels more dangerous now. It always was a capacity protection mechanism — a way of saying "we can only orbit so close to this mass before we lose control." But pre-AI, scope creep added days. Post-AI, scope creep adds what feels like months of output in days. The gravitational pull of AI-accelerated work makes every additional ask exponentially more expensive in terms of the gap it creates.

Five Scenarios of Time Dilation at Work

Let's make this concrete. Here are five common moments when you step away — and what the dilation looks like:

Section III: The Dilation Clock

Cooper watched his children age on a screen. You watch your Slack unread count climb. Different medium, same physics. Here's a clock that shows the gap opening in real time — your human token output versus what an AI-augmented team produces in the same wall-clock seconds:

That's not a simulation of the future. That's what's happening right now, in every AI-augmented workplace, every time you step away from the keyboard. The clock doesn't stop when you do. It accelerates.

Section IV: Capacity Protection Is a Universal Law

Here's the insight that connects Cooper's problem, Anthropic's problem, and your project manager's problem: they're all solving for the same variable. Finite capacity in the face of infinite (or rapidly scaling) demand.

The solutions are identical, too. Not similar — identical. The same mechanisms appear independently in human organizations, AI systems, and physics because they're not cultural artifacts or engineering patterns. They're mathematical necessities that emerge from any system with bounded resources and unbounded demand.

Look at the symmetry. Every human capacity protection mechanism has an exact analog in how Anthropic manages its infrastructure, and both map to fundamental physics. This isn't coincidence. It's convergent evolution. Any sufficiently complex system with finite capacity independently discovers the same solutions:

Section V: Why Anthropic's Throttle Matters More Than You Think

Most commentary on Anthropic's March 2026 capacity adjustments has focused on user frustration. "I'm paying $200/month and I can't use the tool for a full day." Fair complaint. But the more interesting story is what the throttle reveals about the state of AI infrastructure.

The specific mechanisms Anthropic chose are textbook capacity protection:

What's remarkable isn't that Anthropic is doing this. It's that these are the exact same mechanisms that human organizations have used for decades to manage knowledge worker capacity. Anthropic didn't invent anything. They rediscovered what every operations manager already knew: when demand exceeds capacity, you either degrade quality for everyone or ration access for some. There is no third option. The physics don't allow it.

Section VI: The Exacerbation Loop

Here's where the time dilation compounds. Every capacity protection mechanism you use — asking a question, raising a gap, flagging scope creep — takes you off the surface. You step away from production to coordinate. And while you coordinate, the AI-augmented world keeps producing.

Pre-AI, this was manageable. You flag a gap in Teams. Someone responds in an hour. During that hour, maybe 1,000-2,000 tokens of work happened. You read the thread and move on.

Post-AI, you flag the same gap. During the hour you wait for the answer, 48,000 tokens of work happened. The prototype you were questioning? It's been built, tested, and deployed. Your gap is no longer a question about an approach — it's a question about a shipped feature.

This creates a vicious loop:

This is the gravitational spaghettification of knowledge work. Get too close to the AI-augmented velocity and the tidal forces tear your capacity protection apart. Your questions can't keep up with the answers. Your reviews can't keep up with the output. Your judgment can't keep up with the volume. The very thing that makes you valuable — human oversight, institutional knowledge, domain judgment — gets compressed past its Schwarzschild radius.

Section VII: What Cooper Did Next

In Interstellar, Cooper doesn't solve time dilation. He can't. It's physics. What he does is account for it. He makes decisions knowing the cost. He stops pretending the math doesn't apply. He factors the dilation into every choice: which planet to visit, how long to stay, what he's willing to lose.

That's the only playbook for AI-accelerated work, too. You can't eliminate the time dilation. The gap between human pace and AI pace is going to widen, not shrink. But you can stop pretending it doesn't exist and start building for it:

Don't Let Me Leave, Murph

Cooper's tragedy wasn't that he went to space. It was that he understood the cost and went anyway, because the mission mattered. He knew every hour on Miller's Planet would cost years. He factored it in. He made the call. And then he sat in the ship and cried watching what the math took from him.

The knowledge worker's version is quieter but structurally identical. You know that stepping away from the AI-accelerated flow — to ask a question, to protect scope, to take PTO, to eat lunch, to think — will cost you. The world dilates while you're on the surface. The Slack threads accumulate. The decisions get made. The codebase evolves. And when you come back, you're Cooper watching the videos.

Anthropic learned this week what Cooper learned on Miller's Planet: you cannot serve infinite demand with finite capacity. The GPUs are the Endurance. The tokens are the time. The throttle is the acknowledgment that physics applies even to companies valued at billions of dollars.

And your project manager — the one who says "what's the priority?" and "that's scope creep" and "let's take that offline" — has been solving the same equation since before AI existed. They were always protecting capacity. They were always managing time dilation. They just didn't have the physics metaphor.

Now they do. It's called Gargantua. And it's the black hole at the center of every AI-augmented workplace, bending time around it, making every hour on the surface cost seven years on the ship.

The question isn't whether the dilation exists. It does. The question is whether you'll account for it — like Cooper did — or pretend the math doesn't apply.

The physics always win.